Summary
NVIDIA and Meta plan to expand work on Meta’s Muse models, targeting improvements to training and inference. Meta also intends to expand its use of NVIDIA computing and networking infrastructure.
NVIDIA and Meta said on October 26, 2026, that they will expand their collaboration on Meta’s Muse models and agents. The companies plan to work on optimisations intended to improve training and inference performance. Meta also intends to expand its deployment of NVIDIA computing and networking systems.
The announcement describes a partnership spanning AI models, the infrastructure that runs them and tools for managing agents. It sets out plans and goals rather than reporting measured performance gains.
Optimising Muse and AI agents
Meta’s Muse models and agents already use NVIDIA infrastructure, according to NVIDIA. The companies plan to optimise workflows across the collaboration, with the stated aims of improving performance and total cost of ownership—the overall cost of running the systems, not just the purchase price of hardware.
Training is the process through which a model learns patterns from data. Inference is the use of a trained model to produce outputs in response to inputs. Improvements to either can affect the resources needed to develop or operate AI services. The announcement does not specify which parts of the Muse workflows will change, but its focus on both stages points to work across model development and deployment.
NVIDIA and Meta are also considering further collaboration involving NVIDIA’s OpenShell Reference Architecture. The release describes NVIDIA OpenShell as intended for safe deployment and NVIDIA Sentry as a tool for monitoring agent behaviour. These are possibilities the companies are considering, rather than confirmed parts of the expanded work.
AI agents are software systems designed to work toward tasks, often by using tools or taking actions beyond generating a single response. Managing how agents are deployed and monitoring their behaviour are therefore relevant to operating them at scale. The announcement identifies those areas as potential extensions of the partnership.
Meta plans to expand NVIDIA infrastructure
Meta intends to expand its deployment of NVIDIA infrastructure, including Vera Rubin and Vera CPUs, for next-generation AI agents. NVIDIA says the systems are intended to help deliver performance per watt—the amount of computing performance produced for a given amount of energy.
Meta also plans to deploy NVIDIA Spectrum-X Ethernet networking and BlueField-4, which NVIDIA says will support network efficiency and throughput. Together, the stated plans cover models and agents as well as computing and networking: the systems that run AI workloads and move data between them.
The announcement also says NVIDIA will invest $3 billion in Meta’s equity offering. For the technical partnership, the central commitment is the planned work on Muse and Meta’s intended expansion of NVIDIA infrastructure; the release provides no quantified results from the planned optimisations.